In a virtual private local area network (“LAN) service (“VPLS”) multihoming network where many POPs are connected though VC (Virtual circuits) and many virtual LANs (“VLANs”) are present, there is a flaw in RFC 4761, which does not account for network outages on both failure and restoration stages. VPLS Multihoming is typically used for redundancy. The switches which provide access to the many VLANS are connected to two provider edge (“PE”) routers. In the same point of presence (“POP”), one of the PE routers will act as a primary whereas the other will be backup. Virtual Circuits (“VC”) interconnect POPS at a layer 2 level using VPLS. The VC is setup between primary PEs only. Note that by default, the VC will identify the Primary as PE associated with the lowest site ID within the same POP.
If a primary link breaks between one of the lowest numbered site pair connecting two POPS within the service provider VPLS network, a new VC must be established based on the next lowest numbered site pair until the broken primary link has been repaired. This results in a substantial loss of traffic. When the link is repaired, the VC will automatically switch back the initial lowest numbered sire pair, and that will result in another loss of traffic. Note that restoration of a service is expected not to result in loss of service especially to the “non-affected” site IDs, which is not the case here.
Because of the need to establish the new LSI interface due to outage or restoration of the same outage, data loss may occur due to one or both of “convergence time data loss” and/or “back door MAC learning data loss.”